Context-free and context-sensitive dynamics in recurrent neural networks

Continuous-valued recurrent neural networks can learn mechanisms for processing context-free languages. The dynamics of such networks is usually based on damped oscillation around fixed points in state space and requires that the dynamical components are arranged in certain ways. It is shown that qualitatively similar dynamics with similar constraints hold for a(n)b(n)c(n), a context-sensitive language. The additional difficulty with a(n)b(n)c(n), compared with the context-free language a(n)b(n), consists of 'counting up' and 'counting down' letters simultaneously. The network solution is to oscillate in two principal dimensions, one for counting up and one for counting down. This study focuses on the dynamics employed by the sequential cascaded network, in contrast to the simple recurrent network, and the use of backpropagation through time. Found solutions generalize well beyond training data, however, learning is not reliable. The contribution of this study lies in demonstrating how the dynamics in recurrent neural networks that process context-free languages can also be employed in processing some context-sensitive languages (traditionally thought of as requiring additional computation resources). This continuity of mechanism between language classes contributes to our understanding of neural networks in modelling language learning and processing.

Dynamics of trapped Bose-Einstein condensates: Beyond the mean-field

Since dilute Bose gas condensates were first experimentally produced, the Gross-Pitaevskii equation has been successfully used as a descriptive tool. As a mean-field equation, it cannot by definition predict anything about the many-body quantum statistics of condensate. We show here that there are a class of dynamical systems where it cannot even make successful predictions about the mean-field behavior, starting with the process of evaporative cooling by which condensates are formed. Among others are parametric processes, such as photoassociation and dissociation of atomic and molecular condensates.

Variable effects of larval size on post-metamorphic performance in the field

Larval quality may be capable of explaining much of the variation in the recruitment and subsequent population dynamics of benthic marine invertebrates. Whilst the effects of larval nutritional condition on adult performance have received the most attention, recent work has shown that larval size may also be an important and ubiquitous source of variation in larval quality. We examined the effects of variation in larval size on the post-metamorphic survival and growth of Watersipora subtorquata in 2 very different habitats - experimental substrata and pier pilings. We found strong effects of larval size on colony performance, although these varied among experiments. For colonies on experimental substrata, larval size positively affected adult survival and, initially, growth. However, after 3 wk in the field, there was no relationship between larval size and colony size, possibly because colonies were completely surrounded by newly settled organisms. Larval size also positively affected post-metamorphic growth of colonies on pier pilings, but, surprisingly, colonies that came from larger larvae had lower survival than colonies from smaller larvae. Overall, variation in larval size will strongly affect the recruitment and subsequent performance of adults in this species, although this may vary among different habitats. This study highlights the importance of examining the effects of larval quality on adult performance in as realistic conditions as possible, because of the strong interaction between larval size effects and the environment.

The benefits of polyandry in the free-spawning polychaete Galeolaria caespitosa

In many species, females are thought to benefit from polyandry due to the reduced risks of fertilization by genetically incompatible sperm. However, few studies that have reported such benefits have directly attributed variation in female reproductive success to the interacting effects of males and females at fertilization. In this paper, we determine whether male x female interactions influence fertilization in vitro in the free-spawning, sessile polychaete Galeolaria caespitosa. Furthermore, we determined whether polyandry results in direct fertilization benefits for females by experimentally manipulating the number of males contributing towards staged spawning events. To test for male x female interaction effects we performed an initial experiment that crossed seven males with six females (in all 42 combinations), enabling us to assess fertilization rates for each specific male-female pairing and attribute variation in fertilization success to males, females and their interaction. This initial experiment revealed a strong interaction between males and females at fertilization, confirming that certain male-female combinations were more compatible than others. A second experiment tested the hypothesis that polyandry enhances female reproductive success by exposing each female's eggs to either a single male's sperm (monandry) or the sperm from three males simultaneously (polyandry). We performed this second experiment at two ecologically relevant sperm concentrations. This latter experiment revealed a strong fertilization benefit of polyandry, independent of the effects of sperm concentration (which were also significant). We suggest that these direct fertilization gains arising from polyandry will constitute an important source of selection on females to mate multiply in nature.

Two-dimensional approximation for tidal dynamics in coastal aquifers: Capillarity correction

Most previous investigations on tide-induced watertable fluctuations in coastal aquifers have been based on one-dimensional models that describe the processes in the cross-shore direction alone, assuming negligible along-shore variability. A recent study proposed a two-dimensional approximation for tide-induced watertable fluctuations that took into account coastline variations. Here, we further develop this approximation in two ways, by extending the approximation to second order and by taking into account capillary effects. Our results demonstrate that both effects can markedly influence watertable fluctuations. In particular, with the first-order approximation, the local damping rate of the tidal signal could be subject to sizable errors.

Bernard Michael Bindon - reproductive physiologist, animal scientist, research leader

This paper is a foreword to a series of papers commissioned on 'the impact of science on the beef industry', where the Beef CRC-related collaborative scientific work of Professor Bernard Michael Bindon will be reviewed. These papers will be presented in March 2006, as part of a 'festschrift' to recognise his wider contributions to the Australian livestock industries for over 40 years. Bindon's career involved basic and applied research in many areas of reproductive physiology, genetics, immunology, nutrition, meat science and more recently genomics, in both sheep and cattle. Together with his collaborators, he made large contributions to animal science by improving the knowledge of mechanisms regulating reproductive functions and in elucidating the physiology and genetics of high fecundity livestock. His collaborative studies with many colleagues of the reproductive biology and genetics of the Booroola Merino were amongst the most extensive ever conducted on domestic livestock. He was instrumental in the development of immunological techniques to control ovulation rate and in examining the application of these and other techniques to increase beef cattle reproductive output. This paper tracks his investigations and achievements both within Australia and internationally. In the later stages of his career he was the major influence in attracting a large investment in Cooperative Research Centres for the Australian cattle industry, in which he directed a multi-disciplinary approach to investigate, develop and disseminate science and technology to improve commercial cattle productivity.

Gas-surface interactions in the thermal and sub-thermal regime: A molecular dynamics study

Molecular dynamics simulations are used to study the interaction of low-energy Ar atoms with the Ni(001) surface, Angular scattering distributions, in and out of the plane of incidence, are investigated as a function of incident energy, angles of incidence, crystallographic orientation of the incident beam and surface temperature. The results show a clear transition to the structure scattering regime at around 2 eV. However, at lower energies, two sub-regimes are revealed by the simulations, Far energies up to 250 meV, scattering is mainly diffuse, and significant trapping on the surface is observed, At energies above this level, lobular patterns start to form and trapping decreases with the increase in energy, Generally, there is a weak temperature dependence, but variations in the angle of incidence and/or changes in the crystallographic direction, generate significant changes in the scattering patterns.

Quantum dynamics of an atomic Bose-Einstein condensate in a double-well potential

We consider the quantum dynamics of a neutral atom Bose-Einstein condensate in a double-well potential, including many-body hard-sphere interactions. Using a mean-field factorization we show that the coherent oscillations due to tunneling are suppressed when the number of atoms exceeds a critical value. An exact quantum solution, in a two-mode approximation, shows that the mean-field solution is modulated by a quantum collapse and revival sequence.

Effect of body composition at selection on reproductive development in large white gilts

Fifty-four Large White gilts were used to determine the effect of body composition at selection (145 d of age) on the onset of puberty and subsequent reproductive development until 202 d of age. Gilts were assigned to one of three groups based on their backfat depth at selection: 10 to 12 mm (L), 13 to 15 mm (M), and 16 to 18 mm (F). All of the F gilts, 92% of the M gilts, and 67% of the L gilts reached puberty by slaughter at 202 d of age. Data from a subgroup (first 67% to reach puberty in each group; L = Lp, M = Mp, and F = Fp) was also used. The M (Mp) and F (Fp) gilts reached puberty at 172 d (166 d) and 170 d (166 d) of age, respectively, but the L (Lp) gilts at 184.5 d were 12 d (18 d) older than M(P < .05), Mp(P < .001), and F(P < .01), Fp (P < .001) gilts. The Lp (97.68 kg) and Mp (98.33 kg) gilts were lighter (P < .01) than Fp (108.72 kg) gilts at puberty. There were no differences (P < .05) among the L, M, and F gilts in terms of backfat depth or weight at puberty. The L (Lp) gilts had a mean of 1.16 (1.75) estrous cycles, which was lower (P < .01) than for M (Mp) and (P < .01) F (Fp) gilts, with 1.96 (2.29) and 2.25 (2.33) cycles, respectively. L (Lp) gilts had fewer (P < .05) follicles, 13.14 (12.63), than either M (Mp), 19.08 (18.71), or F (Fp), 18.25 (17.42) gilts. The number of corpora lutea was not influenced (P > .05) by grouping at selection, but Fp gilts had fewer (P < .05) corpora lutea than Mp or Fp gilts. Live weight at slaughter was not influenced (P > .10) by grouping at selection or subgrouping at puberty. The L gilts with a mean of 18.05 mm of backfat at slaughter were leaner (P < .05) than the F (21.66 mm) but not (P > .10) the M gilts (19.41 mm). Subgrouping had no effect. Fat deposition and protein deposition were higher (P < .05) in those animals that attained puberty. We conclude that the rate of fat and protein deposition seems to be one of the determinants of puberty attainment.

Conformational dynamics of thyroid hormones by variable temperature nuclear magnetic resonance: The role of side chain rotations and cisoid/transoid interconversions

H-1 NMR spectra of the thyroid hormone thyroxine recorded at low temperature and high field show splitting into two peaks of the resonance due to the H2,6 protons of the inner (tyrosyl) ring. A single resonance is observed in 600 MHz spectra at temperatures above 185 K. An analysis of the line shape as a function of temperature shows that the coalescence phenomenon is due to an exchange process with a barrier of 37 kJ mol(-1). This is identical to the barrier for coalescence of the H2',6' protons of the outer (phenolic) ring reported previously for the thyroid hormones and their analogues. It is proposed that the separate peaks at low temperature are due to resonances for H2,6 in cisoid and transoid conformers which are populated in approximately equal populations. These two peaks are averaged resonances for the individual H2 and H6 protons. Conversion of cisoid to transoid forms can occur via rotation of either the alanyl side chain or the outer ring, from one face of the inner ring to the other. It is proposed that the latter process is the one responsible for the observed coalescence phenomenon. The barrier to rotation of the alanyl side chain is greater than or equal to 37 kJ mol(-1), which is significantly larger than has previously been reported for Csp(2)-Csp(3) bonds in other Ph-CH2-X systems. The recent crystal structure of a hormone agonist bound to the ligand-binding domain of the rat thyroid hormone receptor (Wagner et al. Nature 1995, 378, 690-697) shows the transoid form to be the bound conformation. The significant energy barrier to cisoid/transoid interconversion determined in the current study combined with the tight fit of the hormone to its receptor suggests that interconversion between the forms cannot occur at the receptor site but that selection for the preferred bound form occurs from the 50% population of the transoid form in solution.

Antiphase dynamics of a chaotic multimode laser

Antiphase dynamics of an optically pumped NH3 bidirectional ring laser under the chaotic, phase-sensitive mode coupling is experimentally observed. Our experimental result suggests strongly that the dynamics is a generic behavior of the laser.

On the dynamics of mechanical milling in a vibratory mill

The dynamics of mechanical milling in a vibratory mill have been studied by means of mechanical vibration, shock measurements, computer simulation and microstructural evolution measurements. Two distinct modes of ball motion during milling, periodic and chaotic vibration, were observed. Mill operation in the regime of periodic vibration, in which each collision provides a constant energy input to milled powders, enabled a quantitative description of the effect of process parameters on system dynamics. An investigation of the effect of process parameters on microstructural development in an austenitic stainless steel showed that the impact force associated with collision events is an important process parameter for characterizing microstructural evolution. (C) 1997 Elsevier Science S.A.

Structure optimization combining soft-core interaction functions, the diffusion equation method, and molecular dynamics

Smoothing the potential energy surface for structure optimization is a general and commonly applied strategy. We propose a combination of soft-core potential energy functions and a variation of the diffusion equation method to smooth potential energy surfaces, which is applicable to complex systems such as protein structures; The performance of the method was demonstrated by comparison with simulated annealing using the refinement of the undecapeptide Cyclosporin A as a test case. Simulations were repeated many times using different initial conditions and structures since the methods are heuristic and results are only meaningful in a statistical sense.

Non-equilibrium energy and momentum accommodation coefficients of Ar atoms scattered from Ni(001) in the thermal regime: A molecular dynamics study

Molecular dynamics simulations are used to study energy and momentum transfer of low-energy Ar atoms scattered from the Ni(001) surface. The investigation concentrates on the dependence of these processes on incident energy, angles of incidence and surface temperature. Energy transfer exhibits a strong dependence on the surface temperature, at incident energies below 500 meV, and incident angles close to specular incidence. Above 500 meV, the surface temperature dependence vanishes, and a limiting value in the amount of energy transferred to the surface is attained. Momentum exchange is investigated in terms of tangential and normal components. Both components exhibit a weak surface temperature dependence, but they have opposite behaviours at all incidence angles. In each component, momentum can be lost or gained following the interaction with the surface. (C) 1997 Elsevier Science B.V.